Pediatric Radiology

https://doi.org/10.1007/s00247-023-05598-6

ORIGINAL ARTICLE

Assessment of lung ventilation of premature infants

with bronchopulmonary dysplasia at 1.5 Tesla using phase‑resolved
functional lung magnetic resonance imaging
J. P. Dyke1,2   · A. Voskrebenzev3,4 · L. K. Blatt5 · J. Vogel‑Claussen3,4 · R. Grimm6 · S. Worgall5 · J. M. Perlman5 ·
A. Kovanlikaya2

Received: 17 June 2022 / Revised: 15 December 2022 / Accepted: 11 January 2023

© The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2023

Abstract
Background  The most common chronic complication of preterm birth is bronchopulmonary dysplasia (BPD), widely referred
to as chronic lung disease of prematurity. All current definitions rely on characterizing the disease based on respiratory sup-
port level and do not provide full understanding of the underlying cardiopulmonary pathophysiology.
Objective  To evaluate a rapid functional lung imaging technique in premature infants and to quantitate pulmonary ventila-
tion using 1.5 Tesla magnetic resonance imaging (MRI).
Materials and methods  We conducted a prospective MRI study of 12 premature infants in the neonatal intensive care unit
(NICU) using the phase resolved functional lung MRI technique to calculate pulmonary ventilation parameters in preterm
infants with and without BPD grade 0/1 (n = 6) and grade 2/3 (n = 6).
Results  The total ventilation defect percentage showed a significant difference between groups (16.0% IQR (11.0%,18%)
BPD grade 2/3 vs. 8.0% IQR (4.5%,9.0%) BPD grade 0/1, p = 0.01).
Conclusion  Phase-resolved functional lung MRI is feasible for assessment of ventilation defect percentages in preterm infants
and shows regional variation in localized lung function in this population.

Keywords  Neonatal · Lung · Magnetic resonance imaging (MRI) · Ventilation · Phase resolved functional lung ·
Premature · Bronchopulmonary dysplasia

Introduction born at less than 29 weeks’ gestation [1, 2]. Bronchopulmo-

nary dysplasia (BPD) is the major respiratory complication of
Premature infants represent approximately 10% of all births premature birth, affecting 10,000 to 15,000 infants in the USA
in developed countries and preterm birth is associated with a annually, and up to 50% of infants born < 1kg [3, 4]. BPD
mortality rate of up to 24% for extremely premature infants is widely referred to as chronic lung disease of prematurity
and is the most common chronic complication of premature
* J. P. Dyke birth. This disease was first described more than 50 years ago
[email protected] and was characterized by inflammation and fibrosis second-
ary to administration of oxygen and mechanical ventilation in
1
Citigroup Biomedical Imaging Center, Weill Cornell moderately premature infants (~ 34 weeks’ gestation) [1, 5].
Medicine, 1300 York Avenue, Box 234, New York,
NY 10021, USA Mostly due to refinement of clinical management of pre-
2
mature infants, the National Institute of Child Health and
Department of Radiology, Weill Cornell Medicine,
New York, NY, USA
Human Development (NICHD) in 2018 updated the defini-
3
tion of BPD according to the previous criteria and considera-
Hannover Medical School, Hannover, Germany
tion of new treatment strategies [6]. An infant born at less
4
Biomedical Research in Endstage/Obstructive Lung Disease than 32 weeks’ gestation with radiographic-confirmed, per-
(BREATH), German Center for Lung Research (DZL),
Hannover, Germany
sistent parenchymal lung disease and requiring respiratory
5
support at 36 weeks post-menstrual age for more than 3 con-
Department of Pediatrics, Weill Cornell Medicine,
New York, NY, USA
secutive days (to maintain arterial oxygen saturation above
6
90%) would be diagnosed with BPD. This refinement is used
Siemens Healthineers, Erlangen, Germany

13
Vol.:(0123456789)
 Pediatric Radiology

at our institution and considers newer modes of non-invasive to quantify lung ventilation in premature infants with and
respiratory support, and uses grades 1, 2, and 3 instead of without BPD. A benefit of this technique is that the phase-
mild, moderate, or severe [6]. However, there are currently resolved functional lung MRI acquisition uses a routine 2D
no means to non-invasively quantify respiratory function in fast gradient echo multi-phase sequence that may be readily
premature infants other than the degree of respiratory sup- implemented using product sequences on most if not all MRI
port needed to maintain at least 90% oxygen saturation as a scanners. Although implemented on a Siemens MRI scan-
surrogate. Infant pulmonary function testing is not routinely ner, translation to other vendors should occur without issue
used in the clinic as it adds the risk of sedation, cannot be and will be investigated in future work. The sequence should
performed on infants on respiratory support, is not available have the shortest echo time possible (~ 1 ms) and the fastest
in most neonatal units, and lacks the sensitivity to detect temporal resolution for a single or multiple 10 mm thick
early lung disease [7, 8]. coronal slice(s). The processing suite will import digital
Despite the prognostic implications of the current and imaging and communications in medicine (DICOM) images
previous BPD definitions, disease severity is mainly defined from any scanner for analysis. In addition, depending on the
by the degree of respiratory support needed. This does not stability of the infant, the infant may be transported with the
provide a full understanding of the underlying cardiopul- appropriate medical care to an MRI outside the neonatal
monary pathophysiology. The severity of BPD correlates intensive care unit (NICU) which has been done prior to the
with increased in-hospital mortality, morbidity, and the need installation of the MRI in the NICU at our site.
for respiratory support at discharge [1]. The pathogenesis
of BPD is complex and influenced by prenatal and postna-
tal intrinsic and extrinsic factors [9]. Disruption of normal Materials and methods
lung and vascular development, in addition to inflamma-
tory responses induced by infection, mechanical ventilation, Subjects
and oxygen, can result in structural and functional disease
manifestation in airways, lung parenchyma, and vasculature. The study was conducted in the MRI suite contained within
Although the NICHD proposed an update to the grading and the NICU at the Weill Cornell Medicine. Twelve infants
definition of BPD, the confirmation of persistent parenchy- were imaged in the study (4 males/8 females), median post-
mal lung disease, currently relies on chest radiographs along menstrual age at study 40.0  weeks, inter-quartile range
with the need for respiratory support [5]. (IQR)(38.0  weeks, 49.0  weeks), median gestational age
Perinatal and postnatal injury to a premature lung can 29.1 weeks, IQR (26.7 weeks, 30.9 weeks), median weight
affect any number, if not all, of the three main lung struc- at scan 3.1 kg, IQR (2.6 kg, 3.8 kg), median weight at birth
tures, airways, lung parenchyma, and pulmonary vascula- 0.88 kg, IQR (0.65 kg, 1.21 kg), median length 48.8 cm, IQR
ture. Clinical manifestations of BPD airway disease include (45.5 cm, 51.5 cm) (Table 1). Participants were enrolled if
tracheo-bronchomalacia, and increased reactivity of smaller a clinically indicated MRI of the brain was scheduled. One
airways [10]. Parenchymal disease results from disruption additional subject was enrolled and scanned but excluded
of distal lung growth with impaired alveolarization, often due to rotational motion during the phase-resolved functional
with inflammation, resulting in hypercarbia or hypoxemia, lung MRI scan. Additional research imaging of the lung was
which further impairs lung growth [11]. Decreased or abnor- performed following a clinically prescribed brain scan during
mal growth of the pulmonary microvasculature results in the period from May 2021 through March 2022. A separate
pulmonary vascular disease, most commonly manifested as protocol approved by the Institutional Review Board at Weill
pulmonary hypertension [1, 12]. These three disease com- Cornell Medicine with informed consent was obtained from
partments, alone or in combination, often result in signifi- parents of all subjects for MR imaging of the lung.
cant morbidity and an impaired quality of life, e.g., and in
some cases tracheostomy and chronic respiratory support. Magnetic resonance imaging preparation
Stratification of infants with grade 2/3 BPD into subgroups and clinical monitoring
based on their predominant disease phenotype may improve
care for severely affected infants with BPD, as it may allow Infants were prepared for the MRI examination using a “feed
for a more targeted therapeutic approach. There is an urgent and swaddle” technique by nursing staff in the NICU. Infants
need for effective tools and biomarkers to define these BPD were fed and allowed to fall into a natural sleep without
subtypes [13]. sedation or anesthesia. They were provided three layers of
Our study assesses the feasibility of incorporating the hearing protection (soft foam earplugs, plastic ear muffs,
real-time, free breathing, phase-resolved functional lung and blanket padding) and securely swaddled with blankets.
magnetic resonance imaging (MRI) sequence without the A pulse oximeter was placed on the foot and the infant’s
need for additional contrast or sedation as a novel means heart rate and oxygen saturation were monitored through the

13
Pediatric Radiology

Table 1  Subject demographics are given including the degree of home respiratory support needed for each of the premature infants
Subject number Weight (g) Length (cm) Sex PMA at study BW (grams) GA (weeks) NIH 2018 BPD Home
(weeks) definition Respiratory
support

001 2135 45 F 38 1170 32 3/7 N/A† None

002 3275 48 M 38 3/7 2030 32 3/7 N/A† None
003 2833 49 M 38 1460 31 5/7 None None
004 2890 48.5 F 40 890 28 Grade I None
005 3975 51 F 47 6/7 630 29 Grade III Oxygen
006 3952 52 F 49 550 28 3/7 Grade II None
007 3505 49 F 42 610 25 3/7 Grade II Oxygen
008 3730 57.5 F 42 720 25 3/7 Grade II Oxygen
009 2590 44.5 M 37 1200 29 1/7 None None
010 2600 46 F 37 1210 29 1/7 None None
011 5327 58 M 52 665 24 3/7 Grade II None
012 2390 43 F 40 5/7 875 30 1/7 Grade II None

Abbreviations: BPD bronchopulmonary dysplasia, PMA post-menstrual age, BW birth weight, GA gestational age. †N/A (born with a PMA
greater than 32 weeks of age)

duration of the scan. Signals of an infant waking from sleep, blinded to the grade of BPD in these infants for analysis
such as loss of O­ 2 saturation tracing or persistent increase and provided with de-identified DICOM data to maintain
in heart rate, prompted examiners to enter the MRI room to subject privacy. In addition, an interactive prototype appli-
check on the infant immediately. The scan was stopped if the cation, MRLung 2.0 (RG, Siemens Healthcare; Erlangen,
infant did not quickly fall back to sleep [14]. Germany), was used for rapid visualization of the data.
phase-resolved functional lung imaging is a method to evalu-
ate dynamic MRI data acquired in free-breathing without
Magnetic resonance imaging acquisition the need for contrast agent administration. Segmentation of
the lungs was automatically performed with the ability to
All imaging was performed on a 1.5 Tesla (T) MRI scan- manually edit the contour to include additional regions in
ner (MAGNETOM Amira; Siemens Healthcare, Erlangen the periphery (Fig. 1). The acquired image time series was
Germany) located within the NICU using a pair of 8-chan- transformed to a fixed lung volume with a group-oriented
nel flex coils with a coverage of 20 cm × 22 cm (Noras MRI registration approach using code developed in MATLAB
product GmbH; Höchberg, Germany). A single coronal (AV, J-VC) [16]. This enabled the analysis of the signal
slice at the level of the trachea using 2D spoiled gradient time course in each voxel, which was mainly influenced by
echo sequence was acquired with a 168 ms (ms) repeti- signal changes during ventilation (breathing) and perfusion
tion time (TR), 1345 Hz/Pixel receive bandwidth, 1.08 ms (blood flow). During respiration, the MR signal intensity
echo time (TE), 5° flip angle (FA), 10 mm slice thickness, changes between inspiration and expiration. During inspira-
25 cm field of view (FOV) and a 128 × 102 matrix recon- tion, the lungs fill with air and increase in volume reducing
structed to 256 × 256 resulting in an in-plane resolution of the amount of lung tissue in each voxel resulting in a signal
1 mm × 1 mm. Five hundred twelve images were acquired decrease compared with expiration.
continuously over 90 s without respiratory or cardiac gat- Additionally, there is a time-of-flight blood flow effect,
ing. Two additional slices were placed anterior and posterior which describes the inflow of fresh spins with a higher mag-
respectively to the central phase-resolved functional lung netization into the imaging plane such as in MR angiography.
slice and individually acquired under separate series. A Signal will increase and decrease in the time course curve in
reconstruction factor of 25 was applied to get a full range of accord with the cardiac cycle. While the first effect is directly
signal intensities in the lung. connected to ventilation, the second is related to perfusion.
Since these signal changes occur at different frequencies (car-
Magnetic resonance imaging quantitative analysis diac frequency vs respiration frequency), they can be separated
with an adequate filtering approach. Specifically, a low-pass
Image analysis was performed in MATLAB using the filter was applied to evaluate ventilation. Further, an image-
phase-resolved functional lung analysis routine developed guided, edge-preserving filter was applied to increase the
by several of the authors (AV, JVC) [15]. These authors were signal-to-noise ratio while preserving structural details [17].

13
 Pediatric Radiology

Fig. 1  Grade 1 bronchopulmonary dysplasia in a 84-day-old, ex- the lung. b Graph shows change in signal intensity over time with
28-week premature boy. a A single-slice coronal phase resolved func- inspiration and expiration cycles of the free-breathing acquisition on
tional lung magnetic resonance (MR) image of a premature infant the right with a temporal resolution of 168  ms/frame. The frame of
showing automated segmentation of the lungs. Manual editing of the the image on the left shown in the time course is marked by the red
contours was performed to include minor regions at the periphery of line

In the next step, the respiratory phase was estimated for region with high regional ventilation (75–95% quantile).
each image with a sine-model. The sorted image time series Thus, a quantitative correlation map of flow/volume loop in
is interpolated on a regular time-grid to a complete respira- percent was generated: 100% indicating a perfect agreement
tory cycle with 60 phases. This 3D ventilation information with the healthy reference, thus indicative for a normal venti-
(2 spatial dimensions + 1 temporal dimension) can be fur- lation, and 0% with no agreement with the healthy reference,
ther processed to calculate a series of ventilation param- thus indicative for an abnormal ventilation.
eters, including regional ventilation and flow-volume-loop By applying thresholds, binary parameters can be derived
correlation map [18, 19]. The first provides the volume and combined to produce ventilation maps, which can pro-
change during free breathing in relation to the registered vide the percentage of lung volume affected by ventilation
volume, thus in units of ml/ml or %. Since the MR signal defects (Figs. 2, 3, and 4). The total ventilation defect per-
is directly related to volume, regional ventilation can be centage is reported. The ventilation defect percentage is the
derived from regional signals of three respiratory phases: percent of the lung in that slice that is below the 90th per-
centile multiplied by 0.4 for regional ventilation and below
Sref Sref
RVent(x) = − 90% for flow volume loop correlation map with thresholds as
Sinsp Sexp reported previously [25]. The mean regional ventilation and
regional flow-volume score percentages are also reported.
where Sref is the signal midway between inspiration and
expiration, SInsp represents MR signal of the end-inspiratory
state, and Sexp is the MR signal of the end-expiratory state.
Statistical analysis
Regional ventilation dynamics were further analyzed by
Descriptive statistics were computed for all quantitative meas-
calculation of the regional ventilation slopes (ΔRVent/Δt).
ures by reporting the median and interquartile range. A two-
For this purpose, regional ventilation for each respiratory
tailed Mann–Whitney or Wilcoxon Rank Sum non-parametric
phase (Sresp) was calculated:
t-test was performed for each of the ventilation parameters
Sref Sref between groups of BPD grade 0/1 (n = 6) and BPD grade 2/3
RVent(x, t) = − (n = 6) with significance defined as p < 0.05. The non-para-
Sresp Sexp
metric test was chosen given normality could not be assumed
Since regional ventilation is related to volume, the first with this sample size. An online toolbox was used to calculate
derivative of regional ventilation, ΔRVent/Δt is related to the median, inter-quartile range and box plot.
flow and can be used to assess flow/volume loops in analogy
to lung function assessment [20]. To condense the additional
information (flow/volume loop for each voxel) to one number Results
for each voxel, the similarity to a healthy-reference flow/vol-
ume loop of each subject was measured by the cross-correla- Table  2 lists raw MRI results for all subjects separated
tion metric [15]. The healthy-reference flow/volume loop was into those with grade 0/1 BPD vs. grade 2/3 BPD. Table 3
obtained by averaging the flow/volume loops in a selected reports the differences between these two groups. The total

13
Pediatric Radiology

Fig. 2  Grade 1 bronchopulmo-
nary dysplasia in a 84-day-old,
ex-28-week premature boy. a
Anteroposterior chest radio-
graph demonstrates clear lungs.
b A coronal raw regional venti-
lation map prior to thresholding
showing the relative volume
change between expiration and
inspiration in relation to the
registered volume in %. c A
coronal phase resolved func-
tional lung map showing mini-
mal ventilation defect marked
as blue voxels. d A coronal
regional flow volume loop
correlation metric (FVL-CM)
is used as a regional equivalent
to flow-volume lung function
testing by assessment of volume
changes during the whole
respiration cycle. Please note
that homogenous distribution of
FVL-CM > 90% is indicative of
normal ventilation

Fig. 3  Grade 2 bronchopulmonary
dysplasia in a 116-day-old,
ex-25-week premature girl. a
Anteroposterior chest radiograph
demonstrates perihilar opacities
possibly reflecting subsegmental
atelectasis and peribronchial
thickening. b A coronal raw
regional ventilation map prior to
thresholding shows the relative
volume change between expiration
and inspiration in relation to
the registered volume in %. c A
coronal phase resolved functional
lung functional lung map showing
a 13% ventilation defect marked as
blue voxels. d A coronal regional
flow volume loop correlation
metric (FVL-CM) is used as
a regional equivalent to flow-
volume lung function testing by
assessment of volume changes
during the whole respiration cycle.
Please note that homogenous
distribution of FVL-CM > 90% is
indicative of normal ventilation

13
 Pediatric Radiology

Fig. 4  Grade 3 bronchopulmonary dysplasia in a 132-day-old, ex- A phase resolved functional lung functional lung map showing a 30%
29-week premature boy. a Anteroposterior chest radiograph demon- ventilation defect as blue voxels. d A coronal regional flow volume
strates bilateral coarse interstitial and patchy opacities likely repre- loop correlation metric (FVL-CM) is used as a regional equivalent to
senting subsegmental atelectasis. b A coronal raw regional ventilation flow-volume lung function testing by assessment of volume changes
map prior to thresholding shows the relative volume change between during the whole respiration cycle. Please note that homogenous dis-
expiration and inspiration in relation to the registered volume in %. c tribution of FVL-CM > 90% is indicative of normal ventilation

ventilation defect % showed a significant difference between a 30% regional ventilation defect in a 132-day-old, ex-
groups (16.0% IQR (11.0%,18%) BPD grade 2/3 vs. 8.0% 29-week premature boy consistent with grade 3 BPD. The
IQR (4.5%,9.0%) BPD grade 0/1, p = 0.01). Normal ventila- AP chest radiograph in this patient with grade 3 BPD dem-
tion maps and defect maps are shown in an 84-day-old, ex- onstrates bilateral coarse interstitial and patchy opacities
28-week premature boy with grade 1 BPD in Fig. 2. Homo- likely representing subsegmental atelectasis.
geneous parametric maps of regional ventilation and flow/ Ventilation defects may be seen localized to specific
volume loop-correlation map are indicative of grade 0/1 BPD regions of the lung with increasing severity in BPD.
and normal functioning lungs. The anteroposterior (AP) chest Regional ventilation maps prior to thresholding and flow/
radiograph in this patient demonstrates clear lungs. volume loop maps are also shown in each of the infants
Figure 3 shows a 13% regional ventilation defect in a with varying degrees of BPD showing different degrees of
116-day-old, ex-25-week premature girl consistent with homogeneity in agreement with increased severity of BPD.
grade 2 BPD. The AP chest radiograph demonstrates peri- The total ventilation defect difference between grade
hilar opacities possibly reflecting subsegmental atelecta- 0/1 BPD and grade 2/3 is shown in the box plot shown in
sis and peribronchial thickening. Likewise, Fig. 4 shows Fig. 5. All other parameters were similar between groups.

Table 2  Quantitative Subject number

ventilation parameters obtained
from analysis of the phase BPD grade 0/1 001 002 003 004 009 010
resolved functional lung MRI
VDP (total) % 4.5 8.0 9.0 0.0 12.0 8.0
sequence for subjects in each
group. Abbreviations: BPD Mean ventilation % 15.5 5.0 9.0 19.0 11.0 12.0
bronchopulmonary dysplasia, Regional flow-volume score % 96.5 97.0 94.0 99.5 99.0 98.0
VDP ventilation defect Subject number
percentage
BPD grade 2/3 006 007 008 005 011 012
VDP (total) % 18.0 33.0 11.0 11.0 14.0 18.0
Mean ventilation % 10.0 6.0 15.0 6.0 10.0 9.0
Regional flow-volume score % 67.0 98.0 97.0 95.0 97.0 97.0

13
Pediatric Radiology

Table 3  Statistical analysis characterizing the results from Table  2. toddlers [ages 2 days–21 months] were studied with phase-
The median (Q2) and interquartile range (IQR) are reported. The IQR resolved functional lung MRI. Six of the ten subjects were
reports the range from the 25th percentile (Q1) through the 75th per-
centile (Q3). The two-tailed p-values for the Mann–Whitney or Wil-
recruited from the NICU, four were recruited from the
coxon rank sum test between groups are also given. Abbreviations: general pediatric population. Imaging was done at a 3.0
BPD bronchopulmonary dysplasia, VDP ventilation defect percentage Tesla scanner located outside the NICU using sedation in a
BPD 0/1 BPD 2/3 p-val majority of cases [24]. The result for the ventilation defect
percentage was 4.6% ± 7.6% (range 0.0–20.7%) in these
Q2 IQR Q2 IQR
subjects born at term which were similar to our values
VDP (total) % 8.0 (4.5,9.0) 16.0 (11.0,18.0) 0.01 in grade 0/1 BPD preterm infants. Our study adds to this
Mean ventilation % 11.5 (9.0,15.5) 9.5 (6.0,10.0) 0.34 work as we imaged premature infants with and without
Regional flow-volume 97.5 (96.5,99.0) 97.0 (95.0,97.0) 0.38 BPD in the NICU, and we performed the studies without
score % the need for sedation. Our results show that even at a lower
field strength of 1.5 Tesla, that sufficient signal to noise
and temporal resolution for phase-resolved functional lung
analysis can provide potentially sensitive quantified venti-
Discussion
lation parameters of the neonatal lung.
Validation of the phase-resolved functional lung MRI
Our study is the first to assess functional neonatal lung ven-
sequence has also been performed by our co-authors (AV,
tilation defects by quantifying ventilation defect percent-
JVC). The repeatability of the parameters was recently
ages in premature infants with BPD using free breathing
demonstrated in healthy adult subjects and subjects
phase-resolved functional lung MRI with a standard spoiled
with chronic obstructive pulmonary disease (COPD)
gradient echo sequence. Currently, neonatal lung imaging
[25]. Comparisons with gold standard and more estab-
studies routinely focus only on structural abnormalities of
lished techniques have also been carried out. Specifi-
lung parenchyma or tracheal airway using ultra-short echo
cally, phase-resolved functional lung has been validated
time MRI or computed tomography (CT) [21–23]. However,
with fluorinated and hyperpolarized xenon MRI (direct
quantitation of ventilation defects may provide additional
ventilation measurement), dynamic contrast-enhanced
quantitative biomarkers to those measured by structural
(DCE) MRI (perfusion measurement with contrast agent
sequences and to the respiratory support levels. These meas-
application), and standard, single-photon emission com-
ures may also be used to quantitatively assess response to
puterized tomography (SPECT) [26–30]. Correlation
treatment in a non-invasive and rapid manner.
of phase-resolved functional lung imaging results with
This technique has previously been applied primarily
these validation techniques was used to set the optimal
to adults. Recently, lungs of ten healthy term infants and
image thresholds needed to produce the ventilation maps.
Recently, dynamic phase-resolved functional lung param-
eters showed treatment response in COPD [31].
Limitations of our study include the small sample size
and the risk of subject motion during a scan without seda-
tion. Additional subjects in each group may allow more
accurate characterization of lung ventilation in each spe-
cific BPD grade. Rotational motion became evident in one
subject that was excluded from the study in post-process-
ing as the out-of-plane motion cannot be correctly regis-
tered and facilitated a falsely elevated ventilation defect
percentage in a subject without BPD. This necessitated
subsequent viewing of the phase-resolved functional lung
images in real-time to monitor this issue and prevent any
later scans from becoming unusable due to such motion
artifacts. In addition, given we have 512 frames/series,
removal of images with gross motion artifacts may facili-
tate continued analysis of the data. Likewise, as the scans
are only ~ 1.5  min in length, if motion is detected, the
Fig. 5  The total ventilation defect percentage (VDP%) is shown with
scan may immediately be stopped and repeated later in
box and whisker plots comparing infants with grades 0/1 vs. grades
2/3 bronchopulmonary dysplasia (BPD). Statistical significance the exam once the infant has returned to more regular free
(p = 0.01) is shown between groups with minimal overlap breathing. Future work in this area may focus on more

13
 Pediatric Radiology

rapid under-sampled 2D imaging sequences, full 3D acqui- the Netherlands showed only 67% of preterm infants were
sition and more advanced motion correction and filter- properly categorized according to the NICHD workshop
ing of the time intensity curves [32]. Continued work will definition primarily due to a 31% false-negative rate [36].
refine the phase-resolved functional lung defect percentage Our study accurately differentiated premature infants
thresholds specifically for neonates which were originally with grade 0/1 BPD from those with grade 2/3 BPD in an
derived from adults. objective and quantitative manner using phase-resolved
In addition, all research lung scans were performed on functional lung MRI analysis. This classification comple-
infants that had a clinically prescribed MRI brain scan. In ments and confirms that of clinical BPD grading done at
our NICU, it is standard practice that a brain MRI is per- the bedside. Studies in larger cohorts are warranted to
formed on all extremely low birth weight infants before confirm ventilation defect as a predictor for BPD and to
discharge. Additionally, brain MRI may be indicated for confirm agreement with clinical grading of BPD at the
patients with a history of intraventricular hemorrhage, bedside. Additional phase-resolved functional lung MRI
periventricular leukomalacia, ventriculomegaly, neonatal studies may also be used to longitudinally assess func-
seizures, meningitis or other congenital brain malformations. tional changes in lung ventilation in the same infant with
While certain neurological injury may result in respiratory therapeutic interventions.
dysfunction, the majority of patients included in this study
had a normal brain MRI or findings not expected to influence
respiratory drive or mechanics. Conclusions
We have shown that phase-resolved functional lung
MRI is a feasible non-invasive imaging technique in pre- Phase-resolved functional lung lung MRI quantitation is fea-
mature infants performed under free breathing without the sible in preterm neonates with and without BPD, at 1.5 Tesla
need for sedation, or injectable and inhaled contrast agents, without sedation. The technique shows regional variation in
with potential to identify localized defects in ventilation. localized lung function in this vulnerable population and may
This method offers compelling benefits, including dynamic be used to assess ventilation defect percentages. A statistically
regional assessment of lung function (versus the largely significant difference in total ventilation defect percentage was
global assessment of spirometry), and information regard- observed in subjects with BPD grade 2/3 vs. those with grade
ing the entire respiratory cycle, rather than static regional 0/1.
ventilation measures that consider only the inspiration and
Acknowledgements  The authors would like to acknowledge the
expiration time points [23]. Although most MRI vendors sup- assistance of the pediatric and NICU nursing staff at NYP who were
port multinuclear hyperpolarized 129Xenon MRI, upgrading invaluable in the success of this study. Additional appreciation goes to
a clinical scanner requires additional hardware and costly the NYP MRI imaging staff and specifically to Edward Chung, B.S.,
consumables. phase-resolved functional lung can be imple- R.T.(R) (MR)(ARRT), and Yasmin Abdallah, R.T.(R)(MR)(ARRT).
mented with default fast gradient echo sequences provided by Data availability  The datasets generated during and/or analysed dur-
most if not all MRI vendors. Therefore, once a post-process- ing the current study are available from the corresponding author on
ing pipeline is implemented, the method can be applied in a reasonable request.
multi-center or cooperation setting as an outsourced analysis
pipeline without high technological prerequisites. Declarations 
A recent study of 2,677 preterm infants predicted death Conflicts of interest  R.G. is employed by Siemens Healthineers which
or serious respiratory morbidity in 81% of 18–26 month cor- manufactures the MRI scanner and some of the analysis software. All
rected age infants [33]. Identification of exact clinical grade other authors declare no competing interests.
of BPD is crucial as stepwise increase of late death or seri-
ous respiratory morbidity was reported as follows: grade 0
(10%), grade 1 (19%), grade 2 (35%), grade 3 (77%) [32]. References
Note the increase by more than a factor of two in risk from
1. Stoll BJ, Hansen NI, Bell EF et al (2015) Trends in care practices,
grade 0/1 BPD to grade 2/3 BPD. However, accurate bedside
morbidity, and mortality of extremely preterm neonates, 1993–2012.
evaluation of daily oxygen therapy is subjective, based on JAMA 314:1039–1051
the level of experience, can be time consuming and may 2. Northway WH, Rosan RC, Porter DY (1967) Pulmonary disease
also be inconsistently performed [34]. Inter-center variabil- following respiratory therapy of hyaline-membrane disease, bron-
chopulmonary dysplasia. N Engl J Med 276:357–368
ity in assessment technique can also bias the classification
3. Jensen EA, Schmidt B (2014) Epidemiology of bronchopulmonary
of BPD presence and severity. Classification of BPD in the dysplasia. Birth Defects Res A 100:145–157
clinic may also be complicated given supplementary flow 4. Jensen EA, Edwards EM, Greenberg LT et al (2021) Severity of
may be needed to regulate breathing as well as to treat lung bronchopulmonary dysplasia among very preterm infants in the
United States. Pediatrics 148:1–8
parenchymal disease [35]. A database of preterm infants in

13
Pediatric Radiology

5. Doershuk CF, Matthews LW (1969) Airway resistance and lung 25. Pöhler GH, Klimeš F, Behrendt L et al (2021) Repeatability of
volume in the newborn infant. Pediat Res 3:128–134 phase-resolved functional lung (PREFUL)-MRI ventilation and per-
6. Higgins RD, Jobe AH, Koso-Thomas M et al (2018) Bronchopul- fusion parameters in healthy subjects and COPD patients. J Magn
monary dysplasia: Executive summary of a workshop. J Pediatr Reson Imaging 53:915–927
197:300–308 26. Kaireit TF, Gutberlet M, Voskrebenzev A et al (2018) Comparison
7. Davies G (2020) Does newborn screening improve early lung function of quantitative regional ventilation-weighted fourier decomposition
in cystic fibrosis? Paediatr Respir Rev S1526–0542:30121–30124 MRI with dynamic fluorinated gas washout MRI and lung function
8. Shepherd EG, Clouse BJ, Hasenstab KA et al (2018) Infant pulmo- testing in COPD patients. J Magn Reson Imaging 47:1534–1541
nary function testing and phenotypes in severe bronchopulmonary 27. Kaireit TF, Kern A, Voskrebenzev A et al (2021) Flow volume loop
dysplasia. Pediatrics 141:e20173350 and regional ventilation assessment using phase-resolved functional
9. Hayes D Jr, Feola DJ, Murphy BS et al (2010) Pathogenesis of bron- lung MRI: comparison with 129 Xenon ventilation MRI and lung
chopulmonary dysplasia. Respiration 79:425–436 function testing. J Magn Reson Imaging J Magn Reson Imaging
10. Amin RS, Rutter MJ (2015) Airway disease and management in 53:1092–1105
bronchopulmonary dysplasia. Clin Perinatol 42:857–870 28. Kaireit TF, Voskrebenzev A, Gutberlet M et al (2019) Comparison
11. Moschino L, Bonadies L, Baraldi E (2021) Lung growth and pul- of quantitative regional perfusion-weighted phase resolved func-
monary function after prematurity and bronchopulmonary dysplasia. tional lung MRI with dynamic gadolinium-enhanced regional pul-
Pediatr Pulmonol 56:3499–3508 monary perfusion MRI in COPD patients. J Magn Reson Imaging
12. Malloy KW, Austin ED (2021) Pulmonary hypertension in the child 49:1122–1132
with bronchopulmonary dysplasia. Pediatr Pulmonol 56:3546–3556 29. Behrendt L, Voskrebenzev A, Klimeš F et al (2020) Validation
13. Mandell E, Hysinger EB, McGrath-Morrow S (2020) Disease phe- of automated perfusion-weighted phase-resolved functional lung
notyping of infants with severe bronchopulmonary dysplasia. Am J (PREFUL)-MRI in patients with pulmonary diseases. J Magn Reson
Resp Crit Care Med 201:1327–1329 Imaging 52:103–114
14. Dyke JP, Garfinkel AC, Groves AM et al (2018) High resolution 30. Couch MJ, Munidasa S, Rayment JH et al (2021) Comparison of
rapid neonatal whole body composition using 3.0 tesla chemical functional free-breathing pulmonary 1H and Hyperpolarized 129Xe
shift magnetic resonance imaging. Pediatr Res 83:638–644 magnetic resonance imaging in pediatric cystic fibrosis. Acad Radiol
15. Voskrebenzev A, Gutberlet M, Klimes F et al (2018) Feasibility of 28:209–218
quantitative regional ventilation and perfusion mapping With phase- 31 Voskrebenzev A, Kaireit TF, Klimes F et al (2022) Phase-resolved
resolved functional lung MRI in healthy volunteers and COPD, functional lung MRI depicts bronchodilator changes in COPD: A
CTEPH, and CF patients. MRM 79:2306–2314 retrospective analysis of a randomized controlled trial. Radiol: Car-
16. Voskrebenzev A, Gutberlet M, Kaireit TF et al (2017) Low-pass diothora Imaging 4(2):e210147
imaging of dynamic acquisitions (LIDA) with a group-oriented 32. Klimeš F, Voskrebenzev A, Gutberlet M et al (2021) 3D phase-
registration (GOREG) for proton MR imaging of lung ventilation. resolved functional lung ventilation MR imaging in healthy volun-
Magn Reson Med 78:1496–1505 teers and patients with chronic pulmonary disease. Magn Reson Med
17. He K, Sun J, Tang X (2013) Guided image filtering. IEEE Trans 85:912–925
Pattern Anal Mach Intell 35:1397–1409 33. Jensen EA, Dysart K, Gantz MG et al (2019) The diagnosis of bron-
18. Klimeš F, Voskrebenzev A, Gutberlet M et al (2019) Free-breath- chopulmonary dysplasia in very preterm infants. An evidence-based
ing quantification of regional ventilation derived by phase-resolved approach. Am J Respir Crit Care Med 200:751–759
functional lung MRI. NMR Biomed 32:e4088 34. Hines D, Modi N, Lee SK et al (2017) Scoping review shows wide
19. Alsady TM, Voskrebenzev A, Greer M et al (2019) MRI-derived variation in the definitions of bronchopulmonary dysplasia in pre-
regional flow-volume loop parameters detect early-stage chronic term infants and calls for a consensus. Acta Paediatr 106:366–374
lung allograft dysfunction. J Magn Reson Imaging 50:1873–1882 35. Thebaud B, Goss KN, Laughon M et al (2019) Bronchopulmonary
20. Brookes GB, Fairfax AJ (1982) Chronic upper airway obstruction: Dysplasia Nat Rev Dis Primers 5:78
value of the flow volume loop examination in assessment and man- 36. van Rossem MC, van de Loo M, Laan BJ et al (2015) Accuracy of
agement. J R Soc Med 75:425–434 the diagnosis of bronchopulmonary dysplasia in a referral based
21. Higano NS, Bates AJ, Gunatilaka CC et al (2022) Bronchopulmonary health care system. J Pediatr 167:540–544
dysplasia from chest radiographs to magnetic resonance imaging and
computed tomography: adding value. Ped Radiol 52:643–660 Publisher's note Springer Nature remains neutral with regard to
22. Hysinger EB, Higano NS, Critser PJ et al (2022) Imaging in neonatal jurisdictional claims in published maps and institutional affiliations.
respiratory disease. Paed Respir Rev 43:44–52
23. Higano NS, Ruoss JL, Woods JC (2021) Modern pulmonary imag- Springer Nature or its licensor (e.g. a society or other partner) holds
ing of bronchopulmonary dysplasia. J Perinatol 41:707–717 exclusive rights to this article under a publishing agreement with the
24. Zanette B, Schrauben EM, Munidasa S et al (2022) Clinical feasibil- author(s) or other rightsholder(s); author self-archiving of the accepted
ity of structural and functional MRI in free-breathing neonates and manuscript version of this article is solely governed by the terms of
infants. J Magn Reson Imaging 55:1696–1707 such publishing agreement and applicable law.

13